?

Log in

No account? Create an account

Previous Entry | Next Entry

Cricket tracking: getting close!

Yesterday I was grumpy about figuring out how to set up a Virtual Machine so as to run BioTracker on this computer.

Thankfully, it wasn't as convoluted as I thought, because there are perfectly good instructions on what to do on the BioTrack User Guide page.

Also, it's a bajillion times easier to just use mEncoder from within the Ubuntu VM, so that everything's in one place. Take that, Cygwin.

I also collected some nighttime footage to see how well BioTrack could handle the lighting conditions.

Daytime setup:
Cricket activity II: Daytime setup

Nighttime setup:
Cricket activity II: Infrared setup

Why is the nighttime setup a bit blurry? I don't know, actually, but I would guess it has something to do with switching on the infrared mode changing the lens configuration.

I collected 100 images (1 image/second-ish) and then used mEncoder to stitch them into a video. It still took Backgrounder about 30 seconds to generate a background file from the 100-frame video. It looks like there have been some good improvements to ModelMaker, although I'll need to take a closer look to see about optimizing the ability to distinguish between individuals.

Here's the start of a test track, run using the "out-of-the-box" BioTrack parameters:
BioTrack test

By the end:
BioTrack test

Notice that, in the lower left, BioTrack accidentally swapped the identities of the two crickets. This is because one of the crickets went and hid underneath the water container, and BioTrack assumed that it jumped locations to overlap with the other cricket. It actually probably isn't that hard to edit the output to correct for this, although it would take some thinking, and problems like this tend to accumulate over time. I'll be trying to minimize these sorts of things to the extent possible.

Also notice that it's going to be tricky to individually mark the crickets with different paint colors. That might be unnecessary, however.

Tags:

Comments

( 5 remarks — Remark )
altamira16
Jun. 11th, 2016 01:03 am (UTC)
Is the thing in the upper right eye shadow for when the crickets get bored and decide to throw a slumber party and put excessive eye make up on one another?
rebeccmeister
Jun. 13th, 2016 08:14 pm (UTC)
Yes!

...someone on Twitter remarked that it would probably take a long time for the crickets to solve that Rubik's cube.

In actuality, it's a printout of a color checker. Photographers and others use them to check the color balance. I figured it would be a good array of colors to test and see what shows up under near-infrared light. Answer? Not much...
randomdreams
Jun. 11th, 2016 02:16 am (UTC)
Man, that is a seriously swank setup. I suspect a lot of other people could use it or use it as a starting point.
rebeccmeister
Jun. 13th, 2016 08:13 pm (UTC)
That's my goal with it. A $200 webcam and a PC, ready, set, go!

A person can pay $20,000 to a company that specializes in behavioral software to get a plug-and-play program that can't handle videos that are at all nonstandard. Or a lesser sum to some other company writing tracking software that will offer a questionable level of software support. But at that point, why not just hack away at it yourself?
randomdreams
Jun. 14th, 2016 12:26 am (UTC)
Most people take the "I'd rather pay someone to do that" route, and for good reason, but if someone does it in a sufficiently solid, documented way that is DIY, it offers financial relief to a whole class of people.
( 5 remarks — Remark )

Latest Month

September 2017
S M T W T F S
     12
3456789
10111213141516
17181920212223
24252627282930

Tags

Powered by LiveJournal.com
Designed by Naoto Kishi